10+ years owning the whole stack: product MVPs for C-suites, platforms serving ~1B requests/month, and now on-device inference and memory for AI agents.
I build AI systems, I build with AI daily, and I know the ecosystem inside-out.
Escribano: local-first work memory for AI agents. On-device VLM/LLM inference (MLX), FTS5/BM25 retrieval, and a native MCP server — nothing leaves the machine.
14:32 · VS Code
Debugging a JWT refresh flow in middleware/auth.ts — repeated 401s.
14:47 · Terminal
Committed the refresh-token fix and pushed to the feature branch.
Multi-model routing across 7 models by cost, latency, and reasoning depth. Constrained subagents with read-only review permissions. Architecture docs written for agents to read. Built Escribano end-to-end this way — zero architectural rewrites during the documented prototype.
Evaluated ~7 agent frameworks and ~50 adjacent tools for client advisory work. Orchestration, memory/RAG boundaries, multi-tenancy, and safe-action patterns — recommendations grounded in trade-offs, not hype.
Live products and ventures I'm actively building and maintaining
Resilient, Sovereign Infrastructure
Cut cloud costs 60% and eliminate vendor dependency. Infrastructure migration for European B2B SaaS. EU data residency. Zero lock-in.
Local-First Memory for AI Agents
On-device VLM/LLM inference, FTS5/BM25 retrieval, and a native MCP server — private work memory your agents can query. Local-first, GDPR-compliant, open source.
Strategy on paper. Getting the team to actually change? That's the hard part. I embed with you to turn AI and cloud-native ambitions into engineering reality — tooling, team habits, and architecture.
Senior engineering judgment without the full-time hire. I help early and growing teams build foundations that won't need tearing down in 18 months — from first production system to scaled infrastructure.
Legacy systems, tight deadlines, or a blocked team. I plug in where you need me — architecture, delivery, or modernization — and leave things better than I found them.
Past Experiences
Adopting LLM agents is an architecture problem before it's a tooling problem. I help teams design orchestration, memory and retrieval boundaries, framework selection, and safe-action patterns — grounded in real advisory work, not hype.
Most teams have AI tools. Few have changed how they work. I help engineering leaders build the culture, workflows, and tooling so developers stop dabbling and start shipping with AI.
Turn ideas into working software fast. I validate concepts with AI-augmented workflows — no full team required.
A senior engineer's eye on your system. I surface performance bottlenecks, cost leaks, and the technical risks you don't know you're carrying.
Staff-level technical leadership for teams tackling complex product, platform, or AI work. I lead architecture decisions, unblock engineers, and leave systems and operating practices stronger than I found them.
Make your stack work as one. I connect your services, APIs, and data sources — cleanly, reliably, and with the next engineer in mind.
Move to the cloud — skip the war stories. I plan and execute migrations with minimal disruption, leaving your team with infrastructure they actually understand.
Interested in working together? Email is the fastest way to reach me.